Intro to R

ShareIT 3rd October 2024

Chris Brownlie

Me

R user since 2018
BW since October 2022
Management Decision Analytics - Employer Consulting
Recently returned from a 4 month sabbatical!

Assumptions

Know what a programming language is
Have heard of Python, maybe heard of R
Don’t know anything else about it

R is 31 years old

flowchart LR
    a[<b>1976</b> <br><br> 'S' first released]
    b[<b>1995</b> <br><br> 'R' first released]
    c[<b>1997</b> <br><br> CRAN founded]
    d[<b>2000</b> <br><br> R v1.0 release]
    e[<b>2012</b> <br><br> Shiny first released]
    f[<b>2018</b> <br><br> I start using R!]
    g[<b>2024</b> <br><br> R v4.4.1]
    a --> b
    b --> c
    c --> d
    d --> e
    e --> f
    f --> g
    classDef default fill:#276DC2,color:#ffffff,fontSize:30px

Statistical programming language
Written in C and Fortran
Primarily functional

R has been around since 1995
It was heavily inspired by the S programming language, invented in 1976
Specifically written for statistical analysis and visualisation - contrasts to Python which was written to be general purpose
(Written in C like Python but also Fortran)
Most popular programming languages are object oriented (either purely OOP or allow/encourage it)
Some of these also allow alternative approaches like functional programming (e.g. C++)
But OOP has been the dominant programming paradigm/method for the last few decades
With OOP, you focus on objects in your program and how you want them to behave/interact with other objects
Functional style where you focus on independent chunks of code (functions)
Which can be treated as any other data object (assign to a value, store in a list, pass as an argument to another function)
R is a functional language.
Not strictly functional as pure functions (always the same outputs for same inputs, and no side effects) not required
But lends itself to a functional style well

Who uses R?

TIOBE Rankings (all devs)

:::{.notes} - Developed by academics, which is probably why it is more popular in academia - Consistently in the top 15-20 most popular programming languages among software developers - Misleading as it is more popular with data scientists (currently no. 6 on PyPL, which looks at how often tutorials for that language are being googled) :::

PyPL rankings (Google searches)

Kaggle rankings (Data Scientists)

How we use R: IDEs

RStudio

VS Code

Positron

How we use R: Package ecosystem

Why we use R: ‘but Python’?

The big debate

Is Python faster than R?

C vs Julia vs Python vs R

A commonly quoted idea is that R is slow, or at least objectively slower than Python
But this is very reductive, naturally they are both slower than e.g. C because they are both a layer of abstraction over it
For the vast majority of operations, the method you choose is more important than whether you use R or Python
This image shows a particular operation (linear search for membership) with different approaches
Generally it comes down to the developers understanding of the language
(and if there are any specialised tools available which remove the need for that understanding)
Worth bearing in mind who R is popular with: i.e. not software developers, so that is often to blame
Popular with people who want to ‘do stuff with data’ and are less concerned with how well it performs
Software best practices and optimisation are often neglected
Generally, aside from some specialised use cases, anything you can do in Python you can do in R
The developer doing the coding will have a bigger impact than the language being used http://adv-r.had.co.nz/Performance.html

Why we use R: strengths

{tidyverse}

RMarkdown/Quarto

Dynamically generated documents
Static websites
Interactive elements

Shiny

Framework for developing web apps
Easy learning curve
Translates R to web stack

MDA’s Data Navigator Shiny App

Particularly until the last year or so (but even still now), I think shiny is the main advantage R has over Python
If you want to create a data-based web application, shiny (R) is clearly ahead in a few categories:
Ease of learning, flexibility of expression (lazy evaluation means its easy to avoid shiny doing too much), minimal effort for maximal gain
Shiny for Python, Plotly Dash, Streamlit etc. are starting to fill the gap
Shiny for R existed since 2012 so has that advantage over the others
What shiny basically does is translate R code into HTML, CSS and Javascript, so R developers don’t need to know them in order to get started
Allows developers to not have to think about client-server interactions, the websockets that handle that communication or how the complexities of front end development
Naturally if you want to build a production/enterprise app you will need to know more but the ease of getting started is a big plus

R from an IT perspective

Under the hood

Interpreted
Single threaded
Memory inefficient

Dependency management

Docker & {renv}
{rix}
Linux vs Windows

R integrates well with Docker
There is an initiative called rocker which provides a variety of Docker images for R
We use Docker in our ShinyProxy infrastructure for hosting and deploying apps
The renv package creates a snapshot of R package dependencies for a project
Combining Docker and renv is probably the best way of maintaining dependencies for a project, especially in production
{rix} is a newer package, built on the Nix framework, which approaches the same problem in a slightly different way
Not used yet, my understanding is Nix is sort of like a combination of the two, bundles everything into the same environment
Docker still the most popular
Something to be aware of as someone in IT or just developing in R
It is common to develop in Windows and deploy to Linux
R packages are precompiled to binary packages for windows, in linux they are generally installed from source
This means on linux we often need various compilers and build tools that aren’t required in development
Other things which cause issues when deploying include: file paths, case sensitivity,

Security considerations

Open source
Packages are checked/built before being accepted to CRAN
Always a risk, weigh the pros and cons

R doesn’t stop you from shooting yourself in the foot, but as long as you don’t aim the gun at your toes and pull the trigger, you won’t have a problem. - Hadley Wickham, Advanced R

R is an open source language, which means anyone can see the source code for it
There are a lot of resources online which discuss the pros and cons of open source and security implications
General consensus is that open source languages are generally as safe as proprietary software
Although this quote is technically talking about something different (dynamic typing in R), I like it and think it applies to security concerns too
Recently there was a ‘vulnerability’ detected in older versions of R
Reading in unknown files of a particular type could lead to arbitrary code execution, many cybersecurity researchers objected to this as it applies to anything any time (don’t read in files you don’t know the origin of) https://www.linuxfoundation.org/blog/open-source-separating-fact-from-fiction#:~:text=%E2%9C%85Fact%3A%20Unlike%20closed%20source,coding%20practices%20when%20integrating%20dependencies.